Perception robot path planning method combined with obstacle detection

Abstract

The invention discloses a sensing robot path planning method combined with obstacle detection, and belongs to the technical field of intelligent traffic, and the method comprises the steps: dividing a road traffic network into a plurality of road traffic subareas according to the congestion characteristics of the road traffic network; configuring a plurality of sensing robots for the plurality of road traffic partitions; performing patrol path planning according to the road network distribution of the plurality of road traffic zones to obtain a plurality of standard patrol paths; with the plurality of standard patrol paths as constraints, in the process of scheduling the plurality of sensing robots to detect traffic obstacles in the plurality of road traffic zones, dynamically updating the plurality of standard patrol paths based on a detection result to obtain a plurality of updated patrol paths; and pre-constructing a path integration cloud, and loading the plurality of updated patrol paths to the path integration cloud by the plurality of sensing robots. The technical problem that in the prior art, path planning lacks dynamic updating, traffic obstacle changes cannot be coped with in real time, and consequently the patrol efficiency of a sensing robot is low is solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a path planning method for a perception robot combined with obstacle detection. Background Art

[0002] With the acceleration of urbanization and the surge in the number of motor vehicles, urban traffic management faces increasingly severe challenges. Traffic congestion not only leads to low travel efficiency, but also aggravates environmental pollution and energy consumption, seriously affecting the quality of life of residents. In response to these problems, intelligent transportation systems have emerged. Among them, perception robots, as an important part of intelligent transportation systems, can perform real-time data collection, obstacle detection and path planning in complex traffic environments, providing strong technical support for traffic management.

[0003] At present, traditional traffic management methods mainly rely on fixed monitoring equipment and manual scheduling, which have problems such as poor real-time performance, insufficient flexibility, and untimely response to emergencies. With the development of robotics and artificial intelligence, perception robots have gradually been introduced into traffic management to dynamically monitor traffic conditions and detect traffic obstacles. However, existing perception robot path planning methods are mostly based on static environment models, which are difficult to cope with dynamic changes in traffic flow and complex and changeable obstacles. Therefore, there is an urgent need for a perception robot path planning method that combines multimodal obstacle detection and dynamic path planning to improve the real-time response capability and management efficiency of intelligent transportation systems. Summary of the invention

[0004] This application provides a perception robot path planning method combined with obstacle detection, aiming to solve the technical problem in the prior art that path planning lacks dynamic updating and cannot respond to changes in traffic obstacles in real time, thus resulting in low patrol efficiency of perception robots.

[0005] In view of the above problems, the present application provides a perception robot path planning method combined with obstacle detection, the method comprising dividing the road traffic network into multiple road traffic zones according to the congestion characteristics of the road traffic network; configuring multiple perception robots for the multiple road traffic zones; performing patrol path planning according to the road network distribution of the multiple road traffic zones to obtain multiple standard patrol paths; in the process of dispatching the multiple perception robots to perform traffic obstacle detection in the multiple road traffic zones with the multiple standard patrol paths as constraints, dynamically updating the multiple standard patrol paths based on the detection results to obtain multiple updated patrol paths; pre-building a path integration cloud, and the multiple perception robots load the multiple updated patrol paths into the path integration cloud.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By adopting a technical solution of dividing traffic zones based on the congestion characteristics of the road traffic network, configuring perception robots, planning standard patrol routes according to the road network distribution, dynamically updating routes based on detection results, and integrating data through a path integration cloud, the technical problem of the lack of dynamic update of path planning in the existing technology and the inability to respond to changes in traffic obstacles in real time, resulting in low patrol efficiency of perception robots, is solved. The technical effect of optimizing the real-time performance of path planning, improving the patrol efficiency of perception robots and the overall coordination capability of traffic management is achieved.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flow chart of a path planning method for a perception robot combined with obstacle detection is provided for an embodiment of the present application;

[0010] Figure 2 A flow chart of obtaining multiple updated patrol paths in a path planning method for a perception robot combined with obstacle detection is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0011] The overall idea of ​​the technical solution provided by this application is as follows:

[0012] The embodiment of the present application provides a perception robot path planning method combined with obstacle detection. By analyzing the congestion characteristics of the road traffic network, it is divided into multiple traffic zones, and a perception robot is configured in each zone to perform patrol tasks. A standard patrol path is planned based on the road network distribution within the zone, and during the robot's traffic obstacle detection process, the patrol path is dynamically updated according to the detection results, and the updated path is uploaded to the path integration cloud for centralized management and optimization. This solution aims to achieve dynamic adjustment of path planning and coordinated optimization of traffic management, improve the patrol efficiency of perception robots and the real-time response capability of traffic networks.

[0013] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0014] Examples, such as Figure 1 As shown, an embodiment of the present application provides a path planning method for a perception robot combined with obstacle detection, the method comprising:

[0015] Step S100: Divide the road traffic network into a plurality of road traffic zones according to the congestion characteristics of the road traffic network.

[0016] Specifically, the road traffic network refers to the traffic system composed of urban roads, streets, intersections, etc. The function of the traffic network is to connect various areas of the city and support the daily transportation of people and goods. Congestion characteristics refer to the road congestion phenomenon caused by excessive traffic flow in certain areas of the traffic network. Congestion characteristics are usually characterized by data such as traffic flow, vehicle density, and average speed. Common congestion characteristics include increased traffic volume during peak hours and vehicle backlogs on specific sections of the road. Road traffic zoning refers to dividing the road traffic network into multiple small areas according to certain rules, and each area can be monitored, managed, and optimized independently. In path planning and obstacle detection, zoning helps to refine management and improve the efficiency of the system.

[0017] First, it is necessary to collect and analyze basic information about the traffic network. By using tools such as traffic flow monitoring systems and intelligent traffic sensors (such as video surveillance, geomagnetic sensors, and RFID devices), data such as traffic flow, vehicle speed, and road occupancy rate in various parts of the road can be obtained. These data help determine which areas have congestion problems. Based on the collected traffic flow data, the traffic network can be divided into multiple areas with similar traffic conditions through data mining and clustering analysis methods (such as K-means clustering, density clustering, etc.).

[0018] By dividing the road traffic network into multiple traffic zones, it is possible to accurately monitor and manage the traffic conditions in each area. The congestion characteristics of different areas are fully considered, and the patrol robot can adjust the patrol frequency and path according to the real-time traffic flow, thereby improving the efficiency of obstacle detection and traffic management.

[0019] Step S200: configuring multiple perception robots for the multiple road traffic partitions.

[0020] Specifically, a perception robot refers to a robot equipped with a variety of sensors and computing units that can perceive the surrounding environment in real time and perform data processing. Perception robots are usually used to obtain dynamic information about the surrounding environment, especially traffic flow, obstacle locations, road conditions, etc., to support tasks such as path planning and obstacle detection.

[0021] First, according to the size, traffic density, importance and specific needs of the road traffic zone, determine the perception robot that needs to be configured for each zone. Perception robots are usually equipped with a variety of sensors, such as lidar, cameras, millimeter wave radar, ultrasonic sensors, etc. These sensors can help robots identify obstacles, measure distances, and judge road conditions. According to the needs of the zone, select the appropriate type and number of robots for deployment. After the perception robot is configured, each robot needs to be assigned a specific task. Task allocation is generally based on the characteristics of each traffic zone, such as traffic flow, road complexity, and the frequency of obstacles.

[0022] This multi-zone, refined robot configuration ensures that traffic monitoring needs in each area are met. Perception robots can efficiently detect obstacles, collect real-time traffic data, and respond promptly when accidents or traffic jams occur. This deployment method can significantly improve the flexibility and response speed of traffic management, especially when emergencies occur, robots can be quickly dispatched to handle them, avoiding further deterioration of traffic jams.

[0023] Step S300: planning patrol routes according to the road network distribution of the plurality of road traffic zones to obtain a plurality of standard patrol routes.

[0024] Specifically, road network distribution refers to the spatial layout of roads in the road traffic network and their connection relationships. It includes information such as road types (such as highways, urban trunk roads, secondary roads, etc.), road lengths, number of intersections, and traffic flow. The complexity of road network distribution directly affects the difficulty and efficiency of path planning. Patrol path planning refers to setting reasonable moving routes for perception robots so that they can efficiently perform obstacle detection tasks within road traffic zones. Patrol path planning needs to take into account factors such as traffic conditions, obstacles, and road characteristics to ensure that the robot can successfully complete the detection task. The standard patrol path is a standardized path set for the perception robot in the initial planning stage based on the road network distribution and traffic characteristics. These paths serve as reference routes for robot patrols, and are used to guide the robot to successfully pass through each zone and perform obstacle detection during the patrol process.

[0025] First, it is necessary to collect the road network distribution information of each road traffic zone. This includes the type, length, traffic flow, intersection location, common traffic obstacles, etc. of the road. Through tools such as traffic flow monitoring equipment, GPS trajectory data, and map data, detailed road network data for each zone can be obtained.

[0026] Based on the analysis results of the road network distribution, the path planning algorithm (such as Dijkstra algorithm, A* algorithm, genetic algorithm, etc.) is used to determine the standard patrol path. Specifically, the path planning algorithm will consider the following factors: Shortest path: Plan the shortest path from the starting point to the end point in each partition to avoid unnecessary detours. Traffic flow: Consider the congestion of the road and avoid choosing high-traffic areas, especially during peak hours. Obstacle avoidance: By integrating real-time data, dynamically plan a path to bypass traffic obstacles (such as accidents, construction areas, etc.).

[0027] Based on the planning results, multiple standard patrol paths are generated. These paths will serve as the default paths for the perception robot to patrol. Each path contains multiple key points (such as intersections, traffic lights, etc.) and determines the order in which the robot should travel. The planning of each path is usually based on the following goals: cover the entire partition as much as possible, avoid traffic jams, and ensure that the robot can detect and deal with potential obstacles in a timely manner.

[0028] By implementing this step, a reasonable patrol path can be provided for the perception robot in each traffic zone to ensure that it can efficiently complete the obstacle detection task. The optimized standard patrol path can not only avoid traffic congestion areas and obstacles, but also improve the robot's detection efficiency and reduce the running time of invalid paths.

[0029] Step S400: In the process of dispatching the multiple perception robots to detect traffic obstacles in the multiple road traffic zones with the multiple standard patrol paths as constraints, the multiple standard patrol paths are dynamically updated based on the detection results to obtain multiple updated patrol paths.

[0030] Specifically, the updated patrol path is a new patrol path adjusted based on the standard patrol path and actual detection results. The updated path will better adapt to the current traffic conditions, avoid detected obstacles, and improve the safety and efficiency of the path.

[0031] First, in each road traffic zone, the perception robot will patrol according to the preset standard patrol path. These paths are preliminarily designed using path planning algorithms based on factors such as road network distribution and traffic flow.

[0032] During the robot patrol process, various sensors (such as LiDAR, millimeter wave radar, cameras, etc.) will continuously obtain real-time data of the surrounding environment and detect obstacles on the road. These obstacles may make the original patrol path inaccessible or reduce efficiency. 3. Dynamic update of the path.

[0033] Once an obstacle or other emergency is detected, the perception robot will upload the obstacle information to the cloud or central control system, and dynamically update the original path based on real-time data. The update process is usually implemented through a path replanning algorithm, taking into account new road conditions and obstacle locations. To further improve efficiency, multiple perception robots can be coordinated. When a robot completes the path update, the system will synchronously update the patrol paths of other robots to avoid multiple robots avoiding the same obstacle or staying in the same area at the same time. The use of path optimization algorithms (such as ant colony algorithms, particle swarm optimization algorithms, etc.) can effectively improve the coordination of multiple robots, so that the patrol path of the entire traffic zone covers all potential obstacle areas as much as possible, while reducing repeated path patrols.

[0034] Through this step, the perception robot can flexibly adjust the patrol path according to the real-time obstacle detection results, avoiding inefficiency or waste of resources due to changes in traffic conditions. The mechanism of dynamically updating the path improves the responsiveness of traffic management and ensures that the robot can always detect obstacles on the most suitable route. At the same time, the coordinated scheduling and path optimization between multiple robots improves the operating efficiency of the entire system and ensures efficient and comprehensive traffic monitoring.

[0035] Step S500: Pre-constructing a path integration cloud, the multiple perception robots load the multiple updated patrol paths into the path integration cloud.

[0036] Specifically, the path integration cloud refers to a centralized management and optimization platform built on cloud computing technology, which is specifically used to integrate, store, process and optimize data from multiple perception robots or sensors, especially data related to path planning. On this platform, patrol paths generated by multiple robots can be summarized, updated and globally optimized to improve the overall scheduling and path management efficiency.

[0037] During patrol, each perception robot dynamically updates its patrol path based on detected obstacle information, traffic conditions, road condition changes, etc. Through sensors and real-time calculations, the robot generates "updated patrol paths", which are adjustments to the original standard paths, helping the robot avoid new obstacles or respond more effectively to changes in traffic flow.

[0038] Whenever the robot generates a new updated patrol path, these paths are uploaded to the integrated cloud. The cloud platform receives path data from different perception robots and stores and integrates them. The cloud system optimizes and schedules the paths based on factors such as the partitions where different robots are located and the priority of patrol tasks. The path integration cloud is not only responsible for storing and managing updated patrol paths, but also for global path optimization. For example, if some perception robots frequently encounter obstacles in a certain traffic partition, the cloud platform can adjust the patrol path of the entire area based on this information, so that all perception robots can work together efficiently and avoid repeatedly detouring the same obstacle. The cloud system can also predict future traffic conditions and optimize paths in advance based on different traffic conditions, real-time data and historical trends, thereby enhancing the system's response capabilities.

[0039] This step can significantly improve the collaborative operation capability and path planning flexibility of the perception robot system. Through the centralized management of the path integration cloud, the perception robot can share updated patrol paths in real time, improving the real-time and accuracy of path planning. In addition, the path optimization function of the cloud platform can reduce path conflicts between robots and avoid repeated detours around obstacles, thereby effectively improving the patrol efficiency of the entire system and the responsiveness of traffic management.

[0040] Further, such as Figure 2 As shown, in the process of dispatching the multiple perception robots to detect traffic obstacles in the multiple road traffic zones with the multiple standard patrol paths as constraints, the multiple standard patrol paths are dynamically updated based on the detection results to obtain multiple updated patrol paths, and the method also includes: activating the multimodal perception unit of the first perception robot, wherein the multimodal perception unit includes an obstacle information perception module and an obstacle information recognition module; taking the first standard patrol path as a constraint, controlling the first perception robot to collect information within a preset perception scale through the obstacle information perception module during driving in the first road traffic zone to obtain real-time obstacle information; the obstacle information perception module dynamically loads the real-time obstacle information to the obstacle information recognition module to perform traffic obstacle detection and output real-time obstacle risk; according to the real-time obstacle risk, the obstacle avoidance update of the first standard patrol path is performed to output a first obstacle avoidance patrol path; and by analogy, the first standard patrol path is continuously updated according to the obstacle risk output by the multimodal perception unit to output the updated patrol path.

[0041] Specifically, a multimodal perception unit refers to a sensing system that integrates multiple perception technologies, usually including multiple sensors (such as lidar, cameras, millimeter-wave radar, ultrasonic sensors, etc.) to obtain environmental information in different dimensions. The obstacle information perception module refers to a part of the perception unit that is specifically used to obtain obstacle information on the road in real time. It usually includes sensors, signal processing modules, etc., which are used to detect static and dynamic obstacles such as vehicles, pedestrians, and buildings. The obstacle information recognition module refers to another part of the perception system, which is responsible for analyzing, identifying and classifying the raw data collected by the perception module to determine whether it is a traffic obstacle. This module uses image recognition, data fusion and other technologies to identify and analyze obstacles.

[0042] First, during the patrol mission of the perception robot, its multimodal perception unit is activated. This perception unit integrates multiple sensors, such as cameras, lidar, millimeter-wave radar, etc., to perceive the surrounding environment in real time. For example, lidar can provide accurate depth information, cameras can perform image recognition of obstacles, and millimeter-wave radar can penetrate rain and snow to ensure that obstacles can still be effectively detected in bad weather. Through these sensors, the robot can obtain more comprehensive obstacle information.

[0043] The perception robot patrols according to the preset standard patrol path and collects environmental data through the obstacle information perception module. This data includes the location, type, size, speed and other information of traffic obstacles. The obstacle information recognition module processes the collected data, identifies potential obstacles, and calculates the real-time obstacle risk. Based on the real-time obstacle risk, the perception robot will dynamically update the current standard patrol path. If an obstacle is detected in front of the robot, the path planning system will calculate a new obstacle avoidance patrol path based on the location, size and distance of the obstacle to avoid the obstacle and continue patrolling.

[0044] Throughout the patrol process, the perception robot continuously collects new environmental data, evaluates obstacles and risks in real time, and continuously updates the patrol path based on the detection results. As the position of obstacles changes (for example, obstacles move or new obstacles appear), the perception robot will continue to update the standard patrol path to avoid path repetitiveness or potential dangers.

[0045] Through this step, the perception robot can efficiently perform tasks in complex and dynamic traffic environments. It can not only perceive surrounding obstacles in real time, but also adjust the path in real time according to the detection results. This improves the efficiency and safety of traffic obstacle detection and ensures the efficiency and accuracy of traffic management.

[0046] Furthermore, the method also includes: according to the vehicle obstacle avoidance characteristics, obtaining multiple sample unimodal obstacle information through network calls; constructing multiple obstacle identification branches using the multiple sample unimodal obstacle information; connecting the multiple obstacle identification branches in parallel to complete the construction of the obstacle information identification module; after the obstacle information perception module dynamically loads the real-time obstacle information into the obstacle information identification module, the real-time obstacle information is split according to the modality type to obtain multiple unimodal real-time information, and the multiple unimodal real-time information is loaded into the multiple obstacle identification branches to obtain multiple doubtful obstacle risks; interactively verifying the multiple doubtful obstacle risks, and outputting the real-time obstacle risks.

[0047] Specifically, the vehicle obstacle avoidance feature refers to the specific requirements of the vehicle's ability to identify and avoid obstacles during driving, mainly including information such as the height, width, and material of the obstacle. For example, a perception robot may not be able to crush obstacles that are more than 20 cm high or more than 50 cm wide, and needs to bypass them. Sample single-modal obstacle information refers to the obstacle information collected in a specific sensor mode (such as relying only on a single sensor such as lidar, camera, or millimeter-wave radar). The obstacle recognition branch is a branch module in the obstacle recognition system, and each branch specializes in processing obstacle information of a certain modality. The modality type refers to different types of data provided by different perception technologies or sensors. Each modality has different characteristics, and the detection method and recognition results of obstacles are also different. The risk of doubtful obstacles refers to the obstacle risk that the system has not yet fully determined the recognition results of some possible obstacles after processing multimodal data, and there is a certain degree of uncertainty or ambiguity. This risk usually needs to be further confirmed through methods such as cross-validation. Cross-validation refers to the process of mutual verification of multiple modal data or recognition results. Cross-validation of data from different modalities can effectively reduce false recognition and improve recognition accuracy.

[0048] First, based on the vehicle's obstacle avoidance features, the system calls and obtains multiple single-mode obstacle information from different sensors. This information comes from different sensors, and the data collected by each sensor may contain different types of obstacle information. For example, a lidar may provide information about the distance and shape of an obstacle, while a camera provides image information of the obstacle. Through online calls, the system can obtain multiple perception data in real time and enrich its perception of obstacles.

[0049] Using these sample single-modal obstacle information, the system constructs multiple obstacle recognition branches. Each branch is responsible for processing data of a specific modality. These recognition branches operate in parallel, that is, each branch works in parallel and independently processes different perception data. This parallel processing method improves the speed and efficiency of recognition and enhances the system's adaptability to complex environments.

[0050] As the real-time data of the perception system is updated, the obstacle information perception module will dynamically load the real-time obstacle information to the obstacle information recognition module. These real-time information are split according to the modal types of different sensors, that is, each recognition branch receives the information of the corresponding modality for analysis. For example, the lidar branch processes distance information, and the camera branch processes image data. After splitting the real-time obstacle information into multiple single-modal information, each branch receives the data of the corresponding modality for further analysis. After these data are processed by each branch, multiple doubtful obstacle risks may be generated. For doubtful obstacle risks, the system will perform interactive verification, that is, by cross-validating the results of multiple branches, the authenticity of the obstacle is judged. Finally, after the interactive verification process, the system outputs a more accurate real-time obstacle risk assessment result for use by the path planning system.

[0051] This method improves the accuracy and reliability of obstacle detection, especially in complex environments or ambiguous situations, through the interactive verification of parallel obstacle recognition branches and multi-modal perception systems. By dynamically loading real-time obstacle information and cross-validating it, it avoids misidentification of single-modal perception, improves the intelligence and dynamic adaptability of obstacle avoidance paths, and enhances the ability of perception robots to perform tasks in actual road traffic.

[0052] Furthermore, the multiple perception robots load the multiple updated patrol paths to the path integration cloud. Thereafter, the method also includes: the path integration cloud receives a path planning requirement, and the path planning requirement includes a starting position and a target position; the path integration cloud performs path planning in the road traffic network according to the starting position and the target position to obtain multiple alternative driving paths; after the path integration cloud fits the multiple updated patrol paths to the road traffic network, it performs time-consuming calculations on the multiple alternative driving paths in the road traffic network and outputs a target driving path.

[0053] Specifically, path planning requirements refer to the path requirements of users or systems in specific scenarios, including the starting point (departure location) and the end point (target location), as well as any other constraints that may affect path selection (such as traffic conditions, obstacles, etc.). Alternative driving paths refer to multiple candidate paths generated on the road traffic network based on path planning requirements.

[0054] The path integration cloud first receives the path planning requirements, which include the starting location and the target location. These requirements can come from traffic management systems, user terminals or other perception robots. In actual applications, the starting location may be the current location of the robot, and the target location is the area that the robot needs to patrol or detect. Once the path planning requirements are determined, the path integration cloud system uses algorithms (such as A* algorithm, Dijkstra algorithm, etc.) to plan the path in the network based on the information of the road traffic network and calculates multiple alternative driving paths. These alternative paths may consider different goals, such as the shortest path, the least time-consuming path, and the path that avoids peak hours. In order to ensure the feasibility of the path, the system will also check factors such as the state of the road and traffic conditions.

[0055] Next, the path integration cloud will fit the updated patrol paths uploaded by multiple perception robots to the road traffic network. This step is mainly to ensure that the patrol paths generated by the robots can effectively connect with the actual traffic network and conform to the actual conditions of the road. The fitting process includes adjusting the starting and ending points of the path and ensuring that there are no obstacles or traffic restrictions on the roads that the path crosses.

[0056] After the path fitting is completed, the path integration cloud will perform time-consuming calculations on all alternative paths. This calculation not only includes the length of the road, but also takes into account factors such as traffic flow, traffic lights, traffic jams, and the capacity of the road section. Through a comprehensive evaluation of each path, the cloud system can output the optimal target driving path.

[0057] By loading multiple updated patrol paths into the path integration cloud, the system can centrally manage and optimize the planning of multiple robot paths. In this way, the robot can not only dynamically update the path based on real-time data, but also optimize and schedule the global path through the cloud platform. Ultimately, the path integration cloud can output the optimal target driving path, ensuring that the robot can complete the task efficiently and without obstacles, while reducing the risk of traffic congestion and delays, and improving the intelligent management level of the entire transportation system.

[0058] Furthermore, the road traffic network is divided into a plurality of road traffic zones according to the congestion characteristics of the road traffic network, and the method also includes: the interactive GIS system obtains the traffic coverage area of ​​the road traffic network; calls congestion events according to the coverage area of ​​the road traffic network to obtain historical congestion information; and divides the road traffic network into a plurality of road traffic zones according to the historical congestion information.

[0059] Specifically, interactive GIS system refers to a geographic information system that is used to capture, store, analyze, manage and present geospatial data. In this step, "interactive GIS system" refers to a dynamic and operational GIS platform that allows users to interact with geographic data and obtain traffic data in different areas, times and conditions as needed. This system can obtain information such as traffic coverage area, traffic flow, geographic location, etc. in real time. Traffic coverage area refers to the entire road traffic network area covered by GIS system analysis, including various parts of the road, intersections, roundabouts, etc., and displays relevant information such as traffic flow, traffic congestion, and vehicle speed in the area. The coverage area helps to identify key traffic nodes and bottleneck sections. Historical congestion information refers to traffic congestion data in the road network over a period of time in the past. This information includes traffic flow, congestion duration, accident frequency, etc. in different time periods (such as morning peak, evening peak), different roads and areas (such as city center or highway). Historical data helps to identify common congestion patterns, which can then be used to optimize future route planning.

[0060] First, the interactive GIS system is used to load and display the geographic information of the entire road traffic network, including all roads, streets, intersections, traffic nodes, etc. The system generates traffic coverage area maps by connecting real-time traffic sensors and historical data. These area maps show information such as traffic density, road grade, and traffic facilities on each road, which helps the system understand the layout of the entire urban traffic network. Based on the traffic coverage area, the interactive GIS system accesses the historical traffic database and calls up past traffic congestion event data. These data usually come from historical traffic flow monitoring, intelligent traffic signal systems, and accident records. Historical data can show which sections or areas are more prone to traffic congestion during specific time periods (such as holidays, weekday peak hours, etc.) and which areas have relatively stable traffic flow.

[0061] Combining historical congestion information and real-time traffic data, the interactive GIS system can divide the road traffic network into multiple traffic zones. The traffic conditions in each zone are relatively similar, and the system can optimize traffic flow, route planning, etc. based on these zones. In addition, the zone division can also be dynamically adjusted according to the traffic flow in different time periods.

[0062] By dividing the road traffic network into multiple precise traffic zones and analyzing them in combination with historical congestion information, the system can achieve more refined path planning and traffic management. This historical data-based partitioning method helps improve the efficiency and accuracy of path planning, reduce the time wasted by perception robots in congested areas, improve overall traffic monitoring and management capabilities, ensure that robots can perform tasks on the most appropriate paths, optimize patrol routes and reduce the impact of traffic congestion.

[0063] Furthermore, the road traffic network is divided into a plurality of road traffic zones according to the historical congestion information, and the method further includes: locating K road intersections in the road traffic network; obtaining K comprehensive intersection distances by distance calculation based on K groups of adjacent intersections of the K road intersections in the road traffic network; presetting a confidence time domain, and using the confidence time domain as a constraint, obtaining K groups of congestion duration information of the K road intersections from the historical congestion information; and dividing the road traffic network into a plurality of road traffic zones based on the K groups of congestion duration information and the K comprehensive intersection distances.

[0064] Specifically, a road intersection refers to the place where two or more roads intersect, which is often called an intersection or crossroads. The comprehensive intersection distance refers to the distance from a road intersection to other adjacent road intersections. These distances not only take into account the geographical straight-line distance, but can also be comprehensively calculated based on factors such as the specific layout of the road, traffic flow, and common vehicle speeds. The confidence time domain refers to the time range or window set when analyzing historical traffic data, which is used to limit the traffic data within this time range to be considered valid or with a high degree of confidence. Congestion duration information refers to the duration of heavy traffic flow or traffic jams within a specific historical time period.

[0065] First, by locating traffic intersections and calculating the comprehensive intersection distances between them, we can understand the relative positions and traffic density of each intersection. Then, by setting the confidence time domain to filter and call historical congestion duration information, we can obtain accurate historical traffic congestion data. Then, based on this information, we divide road traffic zones to optimize subsequent path planning and traffic management.

[0066] Use GIS system or traffic data analysis tools to locate K important road intersections in the road traffic network. These intersections are key nodes for traffic flow, and they are usually traffic bottlenecks or potential congestion areas. These key intersections can be identified through traffic flow monitoring data. Next, based on the road network layout, calculate the comprehensive intersection distance between these K intersections. This step not only considers the physical distance of the intersection, but also the traffic characteristics of the road, speed limits, common traffic conditions, etc.

[0067] After setting the confidence time domain (such as selecting the peak time period), the congestion duration information of each intersection during this period is obtained by calling historical traffic data. This information can be obtained through the historical records of the traffic monitoring system, traffic flow data collection equipment or traffic management system. For example, there may be a long traffic jam at a certain intersection during the morning peak period (7:00-9:00), resulting in reduced traffic efficiency. According to the K groups of congestion duration information and K comprehensive intersection distances obtained, the road traffic network is divided into multiple traffic zones. The traffic flow, congestion situation, road characteristics, etc. in each zone will be similar, so unified scheduling and optimization can be achieved.

[0068] Through this method, the system can implement more refined traffic management strategies for different areas. In addition, the rational use of historical data and the comprehensive consideration of the distance between intersections can provide more accurate support for future path planning and optimize the operating efficiency of the entire road traffic network.

[0069] Furthermore, the road traffic network is divided into multiple road traffic zones based on the K groups of congestion duration information and K comprehensive intersection distances, and the method also includes: extracting K congestion frequency information based on the K groups of congestion duration information; performing mean calculation on the K groups of congestion duration information to obtain K mean values ​​of congestion clearance time; performing normalized weighted calculation on the K congestion frequency information, the K mean values ​​of congestion clearance time and the K comprehensive intersection distances based on a preset traffic weight set to obtain K traffic congestion coefficients; and using a watershed algorithm to progressively merge road traffic areas according to the K groups of adjacent intersections and the K traffic congestion coefficients to output the multiple road traffic zones.

[0070] Specifically, congestion frequency information refers to the frequency of congestion at a road intersection within a specific time period. It reflects the number of times the intersection has experienced traffic congestion in historical data, which can help determine the traffic pressure and common congestion patterns in the area. The mean time required to clear traffic when congestion occurs at a road intersection. This data is usually calculated based on historical traffic data, taking into account how much time it usually takes to restore normal traffic flow at the intersection when traffic congestion occurs. Traffic weight set refers to a set of weight values ​​used to measure the relative importance of different traffic parameters (such as congestion frequency, clearing time, intersection distance, etc.). In path planning, traffic weight set can help the system adjust the degree of influence of different parameters according to actual conditions to optimize traffic management. Traffic congestion coefficient is an indicator for comprehensively evaluating the traffic conditions of a road intersection, taking into account multiple factors (such as congestion frequency, clearing time, intersection distance, etc.), and is usually calculated by normalized weighting. The watershed algorithm is an algorithm based on image segmentation, which is often used for data clustering and regional division. In this context, the watershed algorithm is used to automatically divide the road traffic network into multiple traffic zones based on the traffic congestion coefficient and the distance between intersections. The algorithm can automatically "merge" related areas based on distance and congestion information, thereby outputting reasonable traffic zone segmentation results.

[0071] First, by analyzing historical traffic data (especially congestion data at intersections), we can extract the congestion frequency information and the average time required to clear the congestion at each intersection. The congestion frequency information reflects the number of times the intersection has been congested in history, helping to determine the common congestion conditions at the intersection; while the average time required to clear the congestion gives the average time required to restore normal traffic flow each time the intersection is congested.

[0072] Next, based on the preset traffic weight set, the congestion frequency information, the average time consumption for clearing traffic, and the distance between intersections of each intersection are normalized and weighted. Normalization is to convert these values ​​of different dimensions (such as congestion frequency, time, and distance) into a unified dimension range (such as between 0 and 1), so as to avoid deviations caused by inconsistent data dimensions. For example, suppose the traffic weight set is: congestion frequency accounts for 40%, clearing time accounts for 30%, and the distance between intersections accounts for 30%. After normalizing these indicators of each intersection, a weighted calculation is performed to obtain the traffic congestion coefficient of each intersection. A higher congestion coefficient indicates that the intersection may face greater traffic pressure.

[0073] Finally, based on the traffic congestion coefficient of each intersection and the comprehensive intersection distance between intersections, the watershed algorithm is used to progressively merge road traffic areas. The watershed algorithm gradually merges road intersections in the network into several traffic partitions by considering the similarity of the distance and congestion coefficient between intersections. The algorithm clusters the severely congested areas together and separates the relatively open areas according to the traffic congestion coefficient of each intersection, thereby forming multiple reasonable traffic partitions. For example, assuming that in a certain traffic network, the traffic congestion coefficients of intersections A and B are high and the distance between them is close, then the watershed algorithm may divide these two intersections into the same traffic partition, and divide intersection C (with a lower traffic congestion coefficient and a longer distance) into another partition.

[0074] Through this method, the system can automatically divide the traffic network into multiple zones based on actual traffic data, providing data support for perception robots to perform tasks such as path planning, obstacle detection, and traffic management in different areas. This method can optimize traffic zones based on historical congestion information and real-time traffic conditions, effectively reduce congestion, improve road utilization efficiency, and provide intelligent support for robot scheduling and traffic management.

[0075] Furthermore, patrol routes are planned based on the road network distribution of the multiple road traffic zones to obtain multiple standard patrol routes. The method also includes: enumerating the passing trajectories of the first road traffic zone to obtain multiple possible passing roads; performing trajectory connection fitting on the multiple possible passing roads to obtain multiple cyclic patrol routes; performing path reliability evaluation based on multiple path jump frequencies and multiple single cycle times of the multiple cyclic patrol routes to locate the first standard patrol route from the multiple cyclic patrol routes; and so on, patrol routes are planned based on the road network distribution of the multiple road traffic zones to obtain the multiple standard patrol routes.

[0076] Specifically, trajectory enumeration refers to enumerating all possible path combinations or driving trajectories in a traffic zone, covering all major roads and intersections in the zone. The goal of trajectory enumeration is to generate all potential patrol paths to ensure that the main traffic nodes of the zone are covered. Trajectory connection fitting refers to connecting multiple enumerated possible roads into a complete path. Connection fitting needs to consider the logical connection between roads, traffic rules (such as one-way streets), and the continuity of the path to generate multiple circular patrol paths. A circular patrol path refers to a path planning mode with the same starting point and end point, and the robot can patrol along this path in a circular manner. The circular path helps to ensure that the roads in the zone are evenly covered while avoiding repeated entry into the same area. The path jump frequency refers to the number of times the perception robot switches from one road to another on a patrol path. The higher the jump frequency, the greater the complexity of the path and the difficulty of connection. The single cycle time refers to the time required for the perception robot to complete a complete circular patrol path. The length of time depends on factors such as the length of the path, traffic conditions, and robot speed.

[0077] In each traffic zone, the passing trajectories are enumerated according to the road network distribution to generate all possible patrol trajectories in the zone. This process usually relies on map data and algorithm tools (such as the traversal algorithm in the figure). Through enumeration, all path combinations covering the main roads of the zone can be listed. After enumerating the possible passing paths, these paths need to be connected and fitted. When fitting, factors such as the continuity, directionality (such as one-way streets), and obstacle distribution of the path need to be considered. Through fitting, multiple complete circular patrol paths can be generated.

[0078] The reliability of the generated loop patrol path is evaluated based on the following two core indicators: Path jump frequency: Paths with lower jump frequency are usually more continuous and suitable for efficient patrols by perception robots. Single cycle time: Paths with shorter time consumption mean higher patrol efficiency. Path reliability evaluation can be achieved through a weighted scoring method. In this way, a reliability score can be calculated for each path, and the path with the highest score can be selected as the standard patrol path.

[0079] For each zone, the path with the highest score is selected as the standard patrol path. The path should have the following characteristics: It can cover the main traffic nodes in the zone. The path jump frequency is low, which reduces the complex operation of the perception robot during driving. The single cycle takes a short time and improves the patrol efficiency. For other traffic zones, the same steps are used to enumerate the passing trajectories, fit the trajectories and evaluate the path reliability, and finally generate a standard patrol path for each zone.

[0080] Through the above method, the system can generate an optimized standard patrol path for each traffic zone. These paths can not only efficiently cover the main roads of the traffic zone, but also reduce the path complexity and patrol time of the perception robot. Ultimately, under the guidance of the standard patrol path, the perception robot can quickly complete traffic obstacle detection and patrol tasks, improve the management efficiency and traffic smoothness of the traffic zone, and reduce resource waste and operating costs.

[0081] In summary, the perception robot path planning method combined with obstacle detection provided in the embodiments of the present application has the following technical effects:

[0082] 1. Realize zoning management and dynamic path updates of road traffic networks, improve patrol efficiency and obstacle detection capabilities of perception robots, and improve overall coordination capabilities and traffic management efficiency through cloud-based integrated paths.

[0083] 2. Obstacle information is collected in real time through the multimodal perception unit to achieve dynamic obstacle avoidance path updates, thereby improving the adaptability and detection accuracy of the perception robot in complex traffic environments.

[0084] 3. Plan standard patrol routes based on road network distribution, determine the optimal route through reliability evaluation, ensure the comprehensiveness and efficiency of patrol coverage, and improve the quality of robot task execution.

[0085] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0086] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A path planning method for a perceptual robot combined with obstacle detection, characterized in that: The method comprises: Dividing the road traffic network into a plurality of road traffic zones according to the congestion characteristics of the road traffic network; configuring a plurality of perception robots for the plurality of road traffic zones; Perform patrol route planning according to the road network distribution of the multiple road traffic zones to obtain multiple standard patrol routes; In the process of dispatching the plurality of perception robots to detect traffic obstacles in the plurality of road traffic zones with the plurality of standard patrol paths as constraints, dynamically updating the plurality of standard patrol paths based on the detection results to obtain a plurality of updated patrol paths; A path integration cloud is pre-built, and the multiple perception robots load the multiple updated patrol paths into the path integration cloud.

2. The path planning method of a perception robot combined with obstacle detection as claimed in claim 1, characterized in that: In the process of dispatching the plurality of perception robots to detect traffic obstacles in the plurality of road traffic zones with the plurality of standard patrol paths as constraints, dynamically updating the plurality of standard patrol paths based on the detection results to obtain a plurality of updated patrol paths, the method further includes: Activating a multimodal perception unit of the first perception robot, wherein the multimodal perception unit includes an obstacle information perception module and an obstacle information recognition module; Taking the first standard patrol path as a constraint, controlling the first perception robot to collect information within a preset perception scale through the obstacle information perception module during driving in the first road traffic zone to obtain real-time obstacle information; The obstacle information perception module dynamically loads the real-time obstacle information to the obstacle information recognition module to perform traffic obstacle detection and output real-time obstacle risk; Performing obstacle avoidance update of the first standard patrol path according to the real-time obstacle risk, and outputting a first obstacle avoidance patrol path; By analogy, the first standard patrol path is continuously updated according to the obstacle risk output by the multimodal perception unit, and the updated patrol path is output.

3. The path planning method of a perception robot combined with obstacle detection as claimed in claim 2, characterized in that: The method further comprises: According to the vehicle's obstacle avoidance characteristics, multiple sample single-mode obstacle information is obtained through network calls; Constructing multiple obstacle recognition branches using the multiple sample single-modal obstacle information; Connecting the plurality of obstacle identification branches in parallel to complete the construction of the obstacle information identification module; After the obstacle information perception module dynamically loads the real-time obstacle information into the obstacle information identification module, the real-time obstacle information is split according to the modality type to obtain a plurality of single-modal real-time information, and the plurality of single-modal real-time information is loaded into the plurality of obstacle identification branches to obtain a plurality of suspected obstacle risks; Interactive verification is performed on the multiple suspected obstacle risks, and the real-time obstacle risks are output.

4. The path planning method of a perception robot combined with obstacle detection as claimed in claim 1, characterized in that: The plurality of perception robots load the plurality of updated patrol paths to the path integration cloud, and then the method further comprises: The path integration cloud receives a path planning requirement, wherein the path planning requirement includes a departure location and a target location; The path integration cloud performs path planning on the road traffic network according to the departure location and the target location to obtain a plurality of alternative driving paths; After the path integration cloud fits the multiple updated patrol paths to the road traffic network, it performs time-consuming calculations on the multiple alternative driving paths on the road traffic network and outputs a target driving path.

5. The path planning method of a perception robot combined with obstacle detection as claimed in claim 1, characterized in that: Dividing the road traffic network into a plurality of road traffic zones according to the congestion characteristics of the road traffic network, the method further comprising: The interactive GIS system obtains the traffic coverage area of ​​the road traffic network; Calling congestion events according to the coverage area of ​​the road traffic network to obtain historical congestion information; The road traffic network is divided into a plurality of road traffic zones according to the historical congestion information.

6. The path planning method of a perception robot combined with obstacle detection as claimed in claim 5, characterized in that: Dividing the road traffic network into a plurality of road traffic zones according to the historical congestion information, the method further comprising: Locating K road intersections on the road traffic network; According to K groups of adjacent intersections of the K road intersections in the road traffic network, distance calculation is performed to obtain K comprehensive intersection distances; Presetting a confidence time domain, and using the confidence time domain as a constraint, obtaining K groups of congestion duration information of the K road intersections from the historical congestion information; The road traffic network is divided into a plurality of road traffic zones based on the K groups of congestion duration information and the K comprehensive intersection distances.

7. The path planning method of a perception robot combined with obstacle detection as claimed in claim 6, characterized in that: Dividing the road traffic network into a plurality of road traffic zones based on the K groups of congestion duration information and the K comprehensive intersection distances, the method further comprising: Extracting and obtaining K congestion frequency information based on the K groups of congestion duration information; Calculate the mean of the K groups of congestion duration information to obtain K mean values ​​of congestion clearing time; Based on a preset traffic weight set, a normalized weighted calculation is performed on the K congestion frequency information, the K congestion clearing time averages and the K comprehensive intersection distances to obtain K traffic congestion coefficients; According to K groups of adjacent intersections and K traffic congestion coefficients, a watershed algorithm is used to progressively merge road traffic areas, and the plurality of road traffic partitions are output.

8. The path planning method of a perception robot combined with obstacle detection as claimed in claim 2, characterized in that: Patrol routes are planned according to the road network distribution of the plurality of road traffic zones to obtain a plurality of standard patrol routes, the method further comprising: Enumerating the passing trajectories of the first road traffic partition to obtain a plurality of possible passing roads; Performing trajectory connection fitting on the multiple possible roads to obtain multiple cyclic patrol paths; performing path reliability evaluation according to multiple path jump frequencies and multiple single-cycle time consumptions of the multiple cyclic patrol paths, so as to locate the first standard patrol path from the multiple cyclic patrol paths; By analogy, patrol route planning is performed according to the road network distribution of the multiple road traffic zones to obtain the multiple standard patrol routes.

Images